Macy's and Sunglass Hut sued for $10M over face-recog arrest and 'sexual assault' A 61-year-old man is suing US retail giant Macy's and the parent biz of chain store Sunglass Hut for $10 million, claiming he was mistakenly arrested in a robbery case after an inaccurate facial-recognition identification match, and subsequently sexually assaulted in jail. In 2022, two men robbed a Sunglass Hut kiosk inside a …
We don't know the precise details of them picking him out though do we? After all the company (their employer) had already "recognised" him using AI and informed the cops. No reason not to think they didn't run the match past the employees beforehand whereby human biases took over. If I was one of the people present during the robbery and you showed me a photo of someone that is, presumably, a close enough match and asked "is this the one" I'd likely confirm it too. Whereas if I were just presented with a selection of similar looking individuals the result may well be different.
And, yes, let's blame AI because too many people consider computers infallible whereas their opinion of people is less, and rightly, so.
Eyewitness recognition has long been known to be unreliable. If you take somebody that looks kinda like the real criminal and ask them if this is the guy, they'll usually answer yes. Present them with a dozen - including a couple ringers - and tell them the guy may or may not be in the lineup, and the witness has a better chance of being right. (But still not great.)
And, according to theory, you can get better results on eye-witness testimony by excluding cases by use of AI -- because AI generally uses different recognition algorithms, it hallucinates orthogonally to human perception.
Of course, neither human nor AI is immune to the problem of "use of old photograph", as is alleged in this case. The odd thing in this case is that the victim was out of custody in a couple of hours -- arrest procedures can take all day, and nobody would have been even considering his alibi in the first couple of hours. In the case of my pasty-pink relative, it took them that long just to decide that he wasn't black.
Depends on how busy they are, the skill of the lawyer involved and motivation.
Consider: The investigators might have a chip on their shoulder in regards to rent-a-cop outfits telling them who the guilty party is and decide to 'prove' he's not the guilty party quickly to show they're far better at their job than some self-appointed-vigilante organisation who thinks they can do better than the professional police. Add in the claimed assault and you've even more motivation for them to dismiss the claims quickly.
Had it been the cops who came up with him being party to the robbery, they'd have more motivation to hold on to him and try to pressure him into proving they were right. In the case of your relative: That's likely why they took their time.
I hope he gets every penny, and then some. AI is fine for drawing pretty pictures and the like, but is not fit for use in applications such as this.
Facial recognition by law enforcement is bad enough, and should be outlawed. But its use by companies in stores is totally unacceptable. A mass boycott of the companies involved, together with an extremely vocal campaign of negative publicity towards them would seem to be in order. In a land of mass consumerism, where the customer is allegedly always right, they need to learn the hard way that this is not acceptable.
This post has been deleted by its author
Yeah, that's something he should be suing for separately.
Even if he had been guilty, there's still no justification for exposing him to that. Jails have a duty to protect their inmates, and failures on this scale should be punished by - at the very least - multiple people losing their careers, if not their liberty.
There is a segment of the population and media which actively glorifies and/or makes a joke of prison abuse, especially rape, as if it were an extra dimension to and a natural consequence of the prison sentence that society imposes. It bleeds through in to the popular culture that normalises the idea. It's fucking abhorrent.
I'm surprised it took him to go into the shop before they decided to incarcerate him.
won't be long before the police will just routinely run cctv through their stores of information from passports and licenses and just apprehend people
this story is why uk police accessing passport databases for scanning & matching photos is a wrong idea
https://www.bbc.co.uk/news/technology-67004576
won't be long before the police will just routinely run cctv through their stores of information from passports and licenses and just apprehend people
If you think that's bad, how about using DNA to predict a face and then using facial recognition on the predicted face. Utter insanity.
Well, apart form the AI generation bit it has already been done. NAZIs, and other authoritarian regimes decided that the shape of a person's head determined what sort of person they were. And August Sander photographed what he considered to be 'types', https://www.tate.org.uk/art/artists/august-sander-5319/five-things-know-august-sander# , although I do not think he had a 'criminal' type, so he was cataloguing rather than predicting.
The Victorians were doing that, it's called Phrenology:
https://en.wikipedia.org/wiki/Phrenology
In fact Terry Pratchett invented for his book series Discworld the idea of retrophrenology - where by the preceise and measured delivery of blows to the head you can give people the traits they desire by reshaping their skull/brain:
https://wiki.lspace.org/Retrophrenology
There's also a linked area of pseudoscience called Craniometry where the size and shape of the skull predicts temprament and traits:
https://en.wikipedia.org/wiki/Craniometry
All hogwash of the highest order.
That reminds me of my office-mate's Clone A Criminal idea - except that it was tongue-in-cheek.
This was before the days of DNA fingerprinting. At that time we used blood grouping including a number of enzymes such as erythrocyte acid phosphatase (EAP) for which there were a number of alleles. In the blood these alleles were expressed as proteins which were sufficiently different to be separated by electrophoresis. They sere still functional enzymes, which allowed for detection, e.g. acid phosphatases will convert mehylumbelliferyl phosphate into a methylumbelliferol which is fluorescent so the locations of the spots on the electrophoresis plate can be seen and compared.
The idea was to be able to retrieve cells from trace tissue such as a hair root, culture them and then get them to express the relevant genes. As I said, strictly tongue-in-cheek - I think.
Where did the presumption of innocence go? And when?
My collection of US movies back to the 30s suggests it was abandoned at least that long ago. In most stories, as soon as a suspect is identified, the cops assume guilt and badger the suspect to 'confess'. As this is so prevalent in the movies, it probably reflects the realities when the movies were made. Rather makes a mockery of "beyond reasonable doubt".
You are still presume innocent until the courts find you otherwise, it's just the Police like to lock you up until we find or make up the evidence to give to the court, as we haven't got the time to go search for those on bail, plus once we have you locked up, we can use physiological pressure, nothing like the threat of a gangrape or the safe solitary cell to get a confession.
Can someone explain to me why the Police used Macy's facial recognition system to 'identify' a suspect? I can understand them asking the witnesses, and looking at Macy's CCTV recordings, but using a private company's 'AI' system to 'guess' the suspect's identity seems to me rather odd, and possibly unprofessional.
The other serious issue with using a facial recognition system is that there tends to be an assumption that it is right, so the Police stop looking for anyone else who might actually be the perpetrator. The gentleman should be suing the police for being too lazy to do their job properly.
Seriously, I hope that entire company goes up in flames. Firstly for unethical shit like this, and secondly for their entirely unethical stranglehold on the prescription glasses market that has made companies get away with charging hundreds of euros for frames that should cost a few bucks apiece, and charging through the nose for the glass that should be also be much much cheaper. That company should never have been allowed to grow to the size it has or get away with some of the business tactics it has employed. (for instance on how it took over Oakley.)
I am open-minded on cases like this as they are usually promoted with a one sided bias. In this case the story is mainly being told through the legal/PR firm representing the suspect.
If the lawyer had provided a specific confirmation on how they could confirm he was 100% in California at the time of the robbery, I would be inclined to believe it 100%. However, if this is a notional "my alibi was I was living in California at the time" that's a bit vague for my liking (preparing an alibi could be easy enough when preparing a violent robbery with guns). If his residency/tenancy/bank records confirm it was impossible for him to be there..then yes sure I can see that alibi holding firm.
Browsing a few other articles on this topic. These articles appear to be promoted by the PR firm employed by his legal team, where the court agreed his alibi was credible enough, but the PR firm are glossing over the suspects long 20 year criminal record (mostly burglary) that put him in the initial crosshairs. With the plaintiff growing up in the area but having moved away in recent years. It's curious that all parties on the opposing side agreed with the match (the company that was robbed, the security staff and AI recognition company, the police, and store staff who identified him in a police line-up, but that is remotely explainable as errors can occur in these circumstances. At that point if the police have agreed with the match and put a warrant out for his arrest - that is tough to blame the AI firm directly.
...this guy was intentionally singled out to be a fall guy for the crime. AI is just as vulnerable to leading questions than people are (if not far more so), and if someone on the facial recog team wanted to "get back" at him for some perceived slight at some point in the past, all they had to do was wait for a case where the perp was "close enough" in appearance, combined with having "outdated photos on file" of the target, to poke and prod the system into hallucinating a match.
The part that's really suspicious though, is the fact that he only spent a few hours behind bars yet immediately got the "soap in a prison shower" experience. Was even THAT part planned? Who else was in on this? How far down does it gooooo...
Either way, as a former Macy's employee, I hope this guy sues for way, way more than that. That entire company was a clown fiesta top to bottom.
The problem is: This facial recognition software will have access to millions of photos. Including old photos. And among millions of photos there will inevitably be someone who looks exactly like the culprit. Or someone who five years ago looked exactly like the culprit looks today. And since they found someone looking exactly like the culprit, the witness will also identify them.
But with millions of photos checked, it’s inevitable that there is a match by pure coincidence. So the match on its own doesn’t mean much.
I guess someone is asking this facial recognition system for the best match. It would be very easy to ask for the 50 best matches and then check whether several of them look very similar. And if you figure out that the three or nineteen best matches all look very similar then we know there is little reason to suspect any of them.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
